Correspondence
Glycosylated Hemoglobin and the Risk of Microalbuminuria in Insulin-Dependent Diabetes Mellitus
N Engl J Med 1995; 333:940-941October 5, 1995
- Article
To the Editor:
The suggestion by Krolewski et al. (May 11 issue)1 that there is a threshold value for glycosylated hemoglobin above which the risk of microalbuminuria abruptly increases has important implications for the treatment of patients with insulin-dependent diabetes mellitus (IDDM). In their review of previous studies, the authors failed to mention the EURODIAB IDDM Complications Study2 involving 3250 patients (twice the size of their study), in which the risk of microalbuminuria increased with worsening glycemic control with no evidence of a threshold,2 even after the patients were grouped according to the duration of IDDM, as was done by Krolewski et al. (Table 1Table 1
Odds Ratios for the Risk of Microalbuminuria According to the Duration of Diabetes and Glycosylated Hemoglobin Values, Adjusted for Age and Sex.).Their assertion of a threshold effect depends almost exclusively on the interpretation of their Figure 1, which shows the relation between hemoglobin A1 and the risk of microalbuminuria, adjusted for the duration of diabetes. But the results in their Table 2 show that the slope of the relation between hemoglobin A1 and microalbuminuria differs as a function of the duration of diabetes; thus, adjusting for the duration of diabetes in the figure is misleading. Instead, the duration-stratified estimates should be plotted. The argument for a threshold effect is driven by the contribution of a small outlying group of patients with extremely poor glycemic control (hemoglobin A1 values, approximately 15 percent). Despite this, the confidence limits around the odds ratios are not presented, making it difficult for readers to judge the validity of the conclusions for themselves. Three models of the relation between hemoglobin A1 and microalbuminuria were tested, but results for only the threshold model are presented. We need to be convinced that the threshold model fits best by seeing the results of the other tests. Producing a plausible biologic post-hoc explanation for a threshold effect does not necessarily prove its existence.
In conclusion, the argument for a threshold in the relation of microalbuminuria to glycemic control is not strong. Patients with IDDM should be encouraged to maintain the best possible level of glycemic control without incurring an unacceptable risk of hypoglycemia.3
3 ReferencesNish Chaturvedi, M.R.C.P.
John H. Fuller, F.R.C.P
University College London, London WC1E 6BT, United Kingdom1
Krolewski AS ,Laffel LMB ,Krolewski M ,Quinn M ,Warram JH . Glycosylated hemoglobin and the risk of microalbuminuria in patients with insulin-dependent diabetes mellitus. N Engl J Med 1995;332:1251-1255
Full Text | Web of Science | Medline2
The EURODIAB IDDM Complications Study
CrossRef | Web of Science | Medline3
The Diabetes Control and Complications Trial Research Group
Full Text | Web of Science | Medline
Author/Editor Response
The authors reply:
To the Editor: The EURODIAB IDDM Complications Study was a survey based on measurements of urinary albumin excretion and hemoglobin A1c on the same day.1 The dose–response relation resulting from this study design is biased toward the null hypothesis. To demonstrate this, we reanalyzed our results using the same design. The results were very similar to those in Table 1 of Chaturvedi and Fuller, with no evidence of a threshold.
Our study differed from the EURODIAB study in two regards.2 First, to overcome measurement variability, we based the classifications of microalbuminuria and hemoglobin A1 on multiple determinations. Second, we measured glycemic exposure during the two years preceding the determination of microalbuminuria.
It is not true that the evidence of a threshold was driven by a “small . . . group of patients with extremely poor glycemic control.” If the 84 patients with persistently high hemoglobin A1 values, represented by the extreme right point in Figure 1 of our paper, are excluded from the analysis, the line above the threshold rises even more steeply. This result is predictable, since that point fell below the fitted line in the figure.
Regarding the three models for the nonlinear relation between hemoglobin A1 and microalbuminuria, note that linear models were rejected by the analyses developed around our Table 2. Thus, the remaining analyses aimed to characterize the nonlinearity rather than to test it repeatedly. The model shown in Figure 1 was selected as the most revealing summary of the data, not as the one with the smallest P value, and interaction terms between the duration of diabetes and hemoglobin A1 were not significant in any of these models.
Interestingly, additional evidence of a threshold can be found in Table 9 of a recent publication from the Diabetes Control and Complications Trial, although the data were not interpreted as such by the authors.3 The risk of microalbuminuria was presented by quartiles of the distribution of base-line hemoglobin A1c values. The median fell somewhat above our threshold but close enough for use as an approximation. In neither the group receiving intensive treatment nor the group receiving conventional treatment was the risk of microalbuminuria lower in the lowest quartile than in the second quartile. Above the median, the risk in the conventional-treatment group rose steeply — from 3.0 cases per 100 person-years to 5.2 and then to 7.1 — as it did above the threshold in our study. Below the median, the risk reduction in the intensive-treatment group was only 0.6 case per 100 person-years (2.4, as compared with 3.0 per 100 person-years in the conventional-treatment group), even though their hemoglobin A1c values averaged approximately 2 percentage points lower during the study. This treatment effect is consistent with the gradual slope below the threshold in our Figure 1. In contrast, in the upper two quartiles the reduction in risk in the intensive-treatment group was 2.7 per 100 person-years, almost five times larger than that below the median.
We disagree with the conclusion of Chaturvedi and Fuller. The evidence for a threshold is substantial. We think, therefore, that patients with IDDM should be encouraged to maintain glycemic control below that threshold, at which the reduction of the risk of renal (and most likely ocular4) complications is substantial but the risk of hypoglycemia is low.
4 ReferencesAndrzej S. Krolewski, M.D., Ph.D.
James H. Warram, M.D., Sc.D.
Joslin Diabetes Center, Boston, MA 022151
The EURODIAB IDDM Complications Study. Microvascular and acute complications in IDDM patients. Diabetologia 1994;37:278-85,
2
Krolewski AS ,Laffel LMB ,Krolewski M ,Quinn M ,Warram JH . Glycosylated hemoglobin and the risk of microalbuminuria in patients with insulin-dependent diabetes mellitus. N Engl J Med 1995;332:1251-1255
Full Text | Web of Science | Medline3
The Diabetes Control and Complications (DCCT) Research Group
CrossRef | Web of Science | Medline4
Warram JH ,Manson JE ,Krolewski AS . Glycosylated hemoglobin and the risk of retinopathy in insulin-dependent diabetes mellitus. N Engl J Med 1995;332:1305-1306
Full Text | Web of Science | Medline
- Citing Articles (3)
Citing Articles
1
Giacomo Deferrari, Maura Ravera, Valeria Berruti. (2003) Treatment of diabetic nephropathy in its early stages. Diabetes/Metabolism Research and Reviews 19:2, 101-114
CrossRef2
Nish Chaturvedi, Simona Bandinelli, Ruggero Mangili, Guiseppe Penno, Raoul E Rottiers, John H Fuller. (2001) Microalbuminuria in type 1 diabetes: Rates, risk factors and glycemic threshold. Kidney International 60:1, 219-227
CrossRef3
Andrzej S. Krolewski, James H. Warram, Maria Beatriz S. Freire. (1996) EPIDEMIOLOGY OF LATE DIABETIC COMPLICATIONS. Endocrinology & Metabolism Clinics of North America 25:2, 217-242
CrossRef
